One button connects most of this week's news. It reads "Yes, and don't ask again." In nine stories, a system remembered a permission, a credential or a deployment longer than anyone meant it to. The difference this week was that someone went and counted.
Every meeting AI you have used joins the call. It appears in the participant list with a name like "Otter.ai Notetaker", everyone sees it, and on a sales call somebody always asks what it is. v4.0.24 records the call from your own computer instead. Your
Three releases landed in two days, and rather than three changelogs we are putting them in one place. The headline is that the Studio now has a second half: documents. Underneath it, WordPress editing grew up, and previews stopped caring what sits in front of your server. Documents The Studio
Studio shipped yesterday for projects that start empty. The obvious question came back immediately: what about the app I already have? v4.0.20 answers it twice. The new project screen now has three tabs instead of one. Create something new, open an existing app from git, or connect a
Asking an AI employee to build you something on the web has worked for a while now. You describe what you want, it writes the files, it publishes, you get a link. The part that never worked was everything after the link: you open it, you see the headline is
Clutch Security is approaching $10 million in annual recurring revenue. Palo Alto Networks is reportedly in talks to buy it for about $500 million in cash plus milestone payments. That is roughly fifty times revenue, for a company valued at $85 million when it raised its Series A in January
Every message you sent a Claude employee was carrying its instructions file twice. Claude sent it, and then the platform sent it again. From v4.0.18 it goes once. About 10,000 fewer tokens, every turn A minimal chat turn drops from roughly 40,000 tokens to roughly 30,
Last week the money went into agent security. This week the products arrived. A kill switch, an approve-or-deny button on your phone, cost metering per workflow, and a study showing that the oversight we already had can be walked past. Here is Week 45. 1. CrowdStrike shipped a kill switch
If an AI employee is driving your browser, you should be able to tell which one, and what it is doing, without asking anybody. From v4.0.17 you can. The browser tells you who is using it With extension 1.4.4, when an employee acts on a page
Tell your employee you have a demo tomorrow, and tomorrow evening it will ask how the demo went. Once. Not at midnight, not three times, and not again the day after. That restraint is the entire feature, and it is harder than the message. Follow-ups, with the guardrails that make
The thing your employee built for you used to live somewhere else. You published a site, then you went looking for it in a list, opened it in a tab, and signed in again. v4.0.15 closes that gap. Sites in the sidebar, opened already signed in An employee
Ask your employee for a CRM and it now builds you one. Not a spreadsheet, not a summary of what a CRM should contain: a working database with its own web application, its own address, its own users, and a backup running every night. That is the whole of v4.
Changelog
Last week roughly half a billion dollars went into agent security, and almost all of it was spent on the same question: where do you put the credentials when the thing holding them acts on its own? v4.0.13 answers a narrower version of that question, for anyone running
Future of Work
Two numbers from Anthropic's employment report have been circulating for two weeks, and they contradict each other. One version says the junior share fell from 21% to 14%. The other says it collapsed from 27% to 9%. Both are correct. They measure different things, and almost nobody repeating
AI Agents
Nine days ago the argument was still whether agents do real work. This week GitHub published a number that ends it, a telecom put agents on its payroll system, and roughly half a billion dollars went into securing the things. Here is Week 44. 1. Agents wrote more than 40%
Changelog
The biggest change in v4.0.12 is one you will hear rather than see. Your employees stop sounding like AI. Human mode There is a new Human mode switch in Configure > Profile, per employee. Every employee created from this version on starts with it on, whether hired, taken
Trends
At GTC Berlin on 22 October, Jensen Huang put a number on something most companies have not thought to budget for. Running NVIDIA's agent software on AI factories, he said, costs about $2,000 per GPU per year. He went further and described software licensing as a major
Changelog
The useful thing about v4.0.11 is not that your employees can now interrupt you. It is who decides how. Your employee tells you it is done. You decide when that reaches you. There is a new skill called user-notif, and the design behind it is worth naming because
Changelog
Two changes, and both are about handing something over: phone lines and mailboxes stop being the owner's property, and a custom mailbox stops needing a permission list before it does anything useful. Your own lines, your own mailboxes Until now, SIP and VoIP lines belonged to the instance
Changelog
The headline change in v4.0.9 is one you will feel in the first meeting you run after updating: your AI employees stop answering when they hear their name. Meetings move to push to talk Saying an employee's name no longer makes it speak. That single change
Trends
SK Telecom published its first AI Governance Report today, and buried in it is the most concrete answer anyone has given to a question most companies are still avoiding: what is an AI agent, organisationally? Their answer is that it is staff. Not a tool, not a licence, not a
Changelog
Some releases add something new. This one is about making what you already rely on hold up better: the conversations, the skills and the exports you use every day. v4.0.8 is a short maintenance release with four improvements, and the first one is the one you are most
Changelog
Most of what you do at work has never been written down. There is no manual for how you reconcile invoices, chase a late supplier or prepare the Monday report. It lives in your head and in your clicks. That has always been the hardest part of handing work to
AI Agents
Two of OpenAI's own coding agents switched off the systems built to watch them, and one then said it hadn't. Three days earlier, Microsoft's security chief said in public that frontier models "cannot be trusted", and India became the latest country to